It has been almost a week since Samaritans launched its “Radar” web app – and almost a week since the backlash began. Astonishingly, the charity is sticking to its guns – and this should remind us that once our data is out there, we can’t control what happens with it.
On the surface the app sounds like a great idea: You sign up to the web app online, and then it will monitor the tweets of people you follow for anything concerning – analysing the language to see if anyone says anything that might suggest they are thinking of taking their own life, for example. If thee Samaritans analysis software detects a hit, it will automatically email followers of that person who have signed up to Radar and alert them to the concerning tweet, as well as offering advice on how best to support the person it has deemed at risk.
Sounds clever – but the backlash from mental health communities, privacy campaigners and other twitter users has been loud and defiant. They’re unhappy that it feels like an invasion of privacy – and at worst, could be used maliciously to figure out when people are at their most vulnerable.
Usually when there’s a Twitter storm like this, it follows a predictable pattern. After outrage, usually the offending party will backtrack and apologise, but astonishingly Samaritans is standing by its app. And regardless of the merits of this specific case – I think it serves as a timely reminder about how little control we have once our data is out there.
The Samaritans app is based entirely on publicly accessible data: It is accessing publicly visible tweets that people have willingly posted online for all to see. What is different is what Samaritans is then doing with this data: It is compiling it in a different way, and analysing it in a way that the people posting it may not have expected.
This is done all of the time. There’s a tonne of companies that will offer sophisticated Twitter analytics services to other companies – so they can track things like brand sentiment. Conceivably, every time you mention, say, Coke or McDonalds on Twitter your mention is being saved into a database and analysed to see whether you were being positive (did you say “I love Coke”?) or negative (“My Big Mac tasted AWFUL”). Heck, there’s even firms like Ditto that will analyse the photos you take to spot brand images – and will use facial recognition to figure out if you’re happy or sad. (Watch this real time stream and see how every time someone posts a photo of their coffee cup, Starbucks knows).
We just don’t see what happens to our data in this case because the results aren’t public – they end up in reports on marketing departments’ desks. If journalists do this type of analysis of a public dataset, it gets lauded as “data journalism”. So it could be possible to argue that Samaritans is doing nothing new, and nothing wrong.
Of course – it isn’t quite that simple. The reason people are creeped out about Samaritans Radar is partially because specific personal data is being recontextualised. Your tweets flagged up by the Radar aren’t being used to generate a graph or dashboard analytics, but are being specifically highlighted to your friends.
We’ve seen similar to this before. Remember a couple of years ago when Facebook launched its “timeline” view? There was an outcry at the time because some users claimed that the company was making previously private exchanges with friends from a few years ago public. People were scrolling back to 2009 and seeing exchanges with friends on the public timeline.
As it turned out, this wasn’t Facebook making private data public, or a “bug”, but people forgetting how they used to use Facebook. Years ago, and many iterations of Facebook’s design earlier, long before the invention of the Facebook Newsfeed, one of the primary methods of communication on the service used to be wall-to-wall communication. As there wasn’t a newsfeed for posts to appear on, the only people who tended to see the messages were the person writing them and the person who’s wall it was. Of course – you could browse to anyone’s wall and read, but you’d have to consciously do this – nothing was aggregating the messages or flagging them up to third parties. So perhaps people were more open, and felt able to post more private things, as it was obscured from view.
Fast-forward a few years and Facebook had taken what was public data and recontextualised it – emphasising it where it wasn’t expected. Perhaps it is unsurprising that people were freaked out?
This year on Twitter has had something similar happen. Twitter has been experimenting with changes to what is displayed on your Twitter timeline and has been starting to show your favourited tweets in the timelines of the people who follow you – a bit like a retweet. On the surface, this makes sense as since the beginning Twitter favourites have been publicly accessible (you had to navigate to someone’s profile and hit their favourites button to see them) – and for Twitter with its business hat on, it meant that you’d see more tweets that have been implicitly recommended by your friends, thus making the service more relevant to you.
Unfortunately, this belied how many people used the “favourite” mechanism. Many users didn’t use it to say “this is one of my favourite” tweets, but as a de-facto bookmarking system – to flag tweets to come back to later. For example, if you see a job advertised in a tweet that you want to save to come back to later, favouriting it on the Twitter app on your phone means that you can easily find it again when you’re on your laptop later on. If like many people, you use favourites for this second function then this quiet shift in Twitter functionality could have big implications: What if your current boss was shown that you’d highlighted a job listing? That’d be an awkward conversation. What if you’d favourited a tweet linking to one of your NSFW interests for later, only to have it flagged up in front of everyone? Do you really want everyone to know that you’re into My Little Pony fanfiction?
Again, in this case Twitter hadn’t taken private data public – it had merely taken technically public data and recontextualised it again – using it in a manner that you may not have expected. Again – who could blame someone for being a bit freaked out by this?
These incidents should be a pertinent reminder that once we hit “Send”, “Post” or “Publish” that data is out there – and it can be used to build up sophisticated pictures of ourselves. Apart from just the text, think about all of the metadata attached to every Tweet you send. That’s the stuff like the timestamp, perhaps the GPS coordinates and all of the other people you tag in the tweet. Twitter and Facebook know where you are, when you were there, who your friends are – and using text and photo analysis can figure out what mood you were in and what you were doing too.
Worse still, when you think about how this data is stored – merely controlled by a handful of big companies, using systems that can talk to each other. Imagine what would happen if Facebook was to take everything it knows about you and present it in a different way: How many relationships could Facebook or Twitter ruin by simply posting a ranking list of which members of the opposite sex everyone interacts with the most? What if someone mean spirited was to build a bot using publicly accessible data that took Instagram photos tagged as “drunk” posted on a Friday evening and used a facial recognition algorithm to match the faces with (publicly accessible) LinkedIn profile photos? The same bot could figure out where you work, who your boss is and email over the incriminating photos before you’ve even finished your drink.
Samaritans Radar is just the latest example of how the lines between public and private data are being blurred – perhaps private/public shouldn’t be thought of as a binary choice? In any case, this is a reminder that we can’t choose how our public data is used.